Download Informed Preferences? The Impact of Unions on

Informed Preferences? The Impact of Unions
on Workers’ Policy Views
Sung Eun Kim∗
Yotam Margalit∗†
Abstract
To what extent, and in what way, do labor unions shape workers’ political preferences? Despite a decline in membership rates, few other organizations have the reach
and ongoing access to such a sizable share of the electorate. A key question that arises
is the degree to which this access translates into political influence. Empirical analyses often find that union members exhibit political views that are distinct from other
workers. But it is unclear whether this is an outcome of the unions’ influence on them,
or whether this reflects self-selection, whereby only workers with certain political preferences join the unions in the first place. We address this debate by combining unique
data from a targeted survey of American workers and a set of inferential strategies that
exploit two sources of variation: differences in the legal choice that workers face when
joining or opting out of unions, and over-time reversal in a union’s policy position.
Focusing on workers’ stance on trade policy, we offer evidence that unions influence
the preferences of their members in a significant and theoretically predictable manner.
Moreover, the analysis indicates that self-selection accounts, at most, for a quarter of
the estimated “union effect”. The results illuminate the political impact of unions and
the broader mechanisms by which citizens form policy preferences.
∗
†
Department of Political Science, Columbia University
Department of Political Science, Tel Aviv University
1
Introduction
To what extent, and in what way, do labor unions shape workers’ political preferences?
The importance of unions is often attributed to their role in advancing the interests of
workers, allowing them to overcome problems of collective action and to generate an effective
“voice” (Freeman and Medoff, 1984). Much has been made, therefore, of the decline in the
power of unions in recent decades and the impact of this trend on the representation of the
disadvantaged and less well-off.
Yet despite a decrease in membership rates, union members still represent very sizable
shares in the electorates of most advanced economies: 26% of all workers in Britain, 36% in
Italy, and over 54% of the workforce in Norway. Even in the U.S, a noted example of shrinking
unionization rates, enlisted union members still account for about 11% of the workforce
(almost fifteen million workers), a conservative figure that excludes non-members covered by
union agreements, nor family members whose livelihoods often depend on a unionized wage
earner (OECD, 2013).1 Clearly, few organizations have the reach and ongoing access to such
a significant share of the electorate as labor unions do. A key question is whether and how
this access translates into political influence.
The importance of this question has gained added impetus in recent years, as scholars
have argued that the decline in union membership contributed to a host of regressive labor
market outcomes, including weakening labor protections in the face of growing trade liberalization, the long shrinkage of the real minimum wage, as well as to the overall rise in income
inequality (Bartels, 2009; Card, Lemieux, and Riddell, 2004; Gosling and Lemieux, 2004;
Hacker and Pierson, 2011). These accounts have centered on a number of causal pathways.
Some have emphasized the diminishing role of unions as representatives of low-skilled workers
in wage negotiations, and unions’ declining ability to oversee sector-wide collective bargain1
Data is from: OECD and J.Visser, ICTWSS database (Institutional Characteristics of Trade Unions,
Wage Setting, State Intervention and Social Pacts, 1960-2010), version 3.0 (http://www.uva-aias.net/).
1
ing. Other accounts have emphasized unions’ reduced ability to influence electoral outcomes
via campaign contributions or turnout mobilization drives (Burns, Francia, and Herrnson,
2000; Leighley and Nagler, 2007; Masters and Delaney, 1984; Neustadtl, 1990; Radcliff and
Davis, 2000; Saltzman, 1987). Notably, these accounts have paid little attention to another
potentially important pathway, namely the influence that unions may exert on shaping the
political preferences of their members. If unions are an institution that directly affects the
policy preferences of a large swath of workers, e.g. by providing policy-relevant information
to its members, the decline in union density could help account for the relatively weak public opposition to a host of regressive policies advanced in recent years. Yet this additional
pathway rests on the assumption that unions exert a meaningful influence on the political
preferences of their members, a key assumption that is to date empirically unsubstantiated.
As a recent study summarizing the state of the research on the topic concluded, “After 60
years of research on American unions, we still lack convincing evidence of whether or how
union membership affects political attitudes” (Ahlquist, Clayton, and Levi, 2014).
This lacuna is due at least in part to a number of empirical challenges that arise in
evaluating the political influence of unions on their members. The first is an absence of
appropriate data. To detect the influence of unions on the political preferences of their
members, one would want to compare the views of union members with other workers who
are otherwise similar in all respects but union membership. Yet standard national surveys
do not include sizable samples of both union members and non-members within the same
industry, arguably the most natural comparison group for this type of analysis. Researchers
are thus limited to comparisons between aggregated groupings of union members employed
across different sectors, to similar groupings of non-members. As a result, one cannot tell
whether differences observed in the political preferences of the two aggregated groups arise
from the effect of union membership itself, or whether the differences simply reflect the
variation in interests of workers employed in different industries. A second empirical challenge
2
arises from the fact that even if one overcame the data availability issue discussed above,
and found that union members in a given industry hold policy positions that differ from
those of their non-unionized counterparts, the interpretation would still be unclear. It may
be that participation in the union itself causes workers to adopt certain preferences (i.e., a
treatment effect), but it is equally plausible that workers who choose to join a union differ
from non-unionized workers in other characteristics that account for their divergent political
stances (i.e., a selection effect).
We address these empirical challenges by combining a unique original survey that includes
large samples of workers in a targeted set of industries together with a set of inferential
strategies that allow us to test the relative strength of the competing explanations. Our
analysis focuses on the policy views of workers regarding trade openness, one of the few salient
political issues on which unions across different industries vary significantly in positions and
strength of preferences. To explore the link between the unions’ stance on trade and the
preferences of their members, we generate a new metric of each union’s policy position that
is based on its lobbying efforts on all trade-related bills in the years preceding the study.
Our findings provide support for the so-called treatment effect of unions; that is, for the
argument that unions exert influence on their members in a clear and systematic manner.
In contrast, the evidence suggests that self-selection into unions accounts, at most, for less
than a quarter of the observed difference between union members and their non-unionized
counterparts. More specifically, the analysis points to the important role of unions as information providers, demonstrating a strong relationship between the intensity of unions’
correspondence with their members, the degree of information that members possess about
the issue at hand, and the degree of alignment between the unions and their members on
the issue.
Exploiting differences across the U.S in the legal choice that workers face in joining
or opting out of unions (a.k.a the ‘Right-to-Work’ law), we show that preferences of union
3
members and non-members are inconsistent with the legal differences in selection mechanisms
into unions. We estimate that union membership accounts for about a 41% increase above the
baseline rate in workers’ likelihood of opposing trade liberalization, an effect comparable to
the effect associated with obtaining a college degree, one of the most studied and established
predictors of trade policy preferences (Scheve and Slaughter, 2001a; Hainmueller and Hiscox,
2006). In addition, we leverage the sudden U-turn in the United Auto Worker’s stance toward
a free trade agreement with Korea and examine the reversal’s impact on its members’ stance.
Using pre- and post-shift data, we show that members indeed became more supportive of
trade expansion following their union’s change of position, while the same shift had no
discernible effect on non-members employed in the same industry.
The article contributes to the research on the political impact of unions, demonstrating
their effect in shaping members’ policy positions. Beyond influencing political outcomes
via Political Action Committees (PACs) and lobbying efforts (Facchini, Mayda, and Mishra,
2011; Masters and Delaney, 2005), we show that unions are also able to influence the views of
their membership in a theoretically predictable and meaningful way. With growing economic
openness and the attendant pressure on labor to be internationally competitive, as well as
the increasing flow of money into politics representing business interests, unions are facing
major challenges that threaten to diminish their effectiveness as workers’ “voice” (Baldwin,
2003; Slaughter, 2007). In these circumstances, assessing and quantifying the influence of
unions on their members’ political preferences is key for mapping the relative forces shaping
today’s political landscape.
More broadly, our analysis also adds to the work on the sources of voter preferences. In
particular, a prominent strand in the political economy literature attributes the positions
that individuals take on various policies – e.g. trade, immigration, taxation – to their
expectations regarding the likely impact of the policy on their wellbeing (Mayda, 2006;
Scheve and Slaughter, 2001a). Yet, most studies typically just assume this link between
4
perceived interests and policy preferences, without explaining how those interests come to
be crystallized by voters. By providing substantial new evidence on the role and impact of
unions as information providers, the paper illuminates one important mechanism that helps
substantiate this key assumption.
The paper proceeds as follows. Section II reviews the main insights from the literature
and draws a set of expectations about the influence of unions on their members. In section
III we describe our data and empirical approach. Sections IV and V present the findings and
a set of robustness tests. The final section discusses the broader implications of the findings
for research on preference formation and the evolving role of unions.
2
Preference Formation, Information, and the Impact of Unions
The determinants of individuals’ policy preferences are a major source of ongoing study.
From the various explanations provided to date, two primary perspectives stand out: one
emphasizing the influence of voters’ self-interested considerations and the other focused on
the impact of ideational factors.
Interest-based explanations suggest that people’s attitudes on a policy are largely determined by the utilities they expect to derive from it. For example, individuals whose
employment is less secure tend to be more supportive of policies that provide generous unemployment insurance or spending on active labor programs (Alesina and La Ferrara, 2005;
Iversen and Soskice, 2001). In the same vein, several studies suggest that individuals exposed to foreign competition in the labor market are more likely to oppose policies liberalizing
immigration or trade (Scheve and Slaughter, 2001a; Mayda and Rodrik, 2005).
Studies emphasizing the role of ideational factors stress the importance of value orientations and partisan attachments in shaping individuals’ policy preferences (Campbell et al.,
1960; Green, Palmquist, and Shickler, 2002; McClosky and Zaller, 1984). For example, beliefs about deservingness and the plight of the poor are shown to affect the level of support
5
for welfare provision (Fong, 2001; Feldman and Steenbergen, 2001). Others show that cosmopolitan inclinations or nationalistic attitudes are closely tied to voters’ preferences on
trade and immigration (Malhotra, Margalit, and Mo, 2013; Margalit, 2012).
Critiques of the interest-based approach center not only on the empirical support (or lack
thereof) for some of its predictions, but also on the mechanism underlying its core logic. In
particular, some question the basic, often implicit, assumption that individuals understand
how their personal well-being is influenced by government policy (Sears and Funk, 1990;
Mansfield and Mutz, 2009).2 The notion that voters can tease out the implications of a complex policy, which at times is a matter of debate even among the experts, seems questionable,
particularly given the wealth of evidence demonstrating citizens’ lack of knowledge or grasp
of very basic political and economic facts (Bennett, 1988; Campbell et al., 1960; Converse,
1962; Ferejohn, 1990; Neuman, 1986).
One response to this critique focuses on voter learning. The process of learning could
presumably occur in several ways, without requiring the (probably, heroic) assumption that
voters actively seek out and process policy-relevant information. For example, individuals
may draw on their everyday experiences to form policy opinions that largely accord with
their interests. Indeed, some studies show that voters update their political preferences
leftwards – even if only temporarily – in response to the experience of various hardships such
as loss of employment or of health care (Hacker, Rehm, and Schlesinger, 2013; Margalit,
2013). Another source of learning is exposure to information or cues. According to this
view, citizens acquire pertinent knowledge about the rationale and preferences of friends,
co-workers, or other groups that they believe to share interests with them, and subsequently
infer how a policy is likely to affect their own interests (Lupia, 1994). It is within this strand
2
Sears and Funk (1990: 164) argue that “ordinary people simply do not often perceive government as
offering them very clear or substantial personal costs or benefits.” Mansfield and Mutz (2009: 432) also
contend that self-interest has only a limited influence on shaping policy preferences because “citizens have a
difficult time linking their personal economic situations to public policies.”
6
of arguments that the importance of unions is often stressed, since unions have close access
to their members via regular meetings, direct mailings or mobilization drives. Unions are
thus able, at least in theory, to communicate to their members facts and opinions over a
range of policy issues. These communications, in turn, could help union members crystalize
their interests and subsequently form or update their preferences (Fordham and Kleinberg,
2012; Leighley and Nagler, 2007).
Indeed, several studies have documented unions’ engagement in a variety of activities
aimed at advancing their policy goals. Such activities include drives to increase voter turnout
among union members and their families (Asher et al., 2001; Becher and Stegmueller, N.d.;
Leighley and Nagler, 2007), efforts to encourage voting for candidates endorsed by unions
(Sousa, 1993; Dark, 1999; Clark and Masters, 2001), mobilization of members to become
more politically active and engage in PAC contribution campaigns, and lobbying activities
aimed at affecting pro-union legislation (Facchini, Mayda, and Mishra, 2011; Freeman and
Medoff, 1984; Masters and Delaney, 1987). Yet, the basic question remains: what do unions
do in terms of shaping the preferences of their own members? Despite the sizable literature
on unions’ operations, there is a striking paucity of evidence regarding the effectiveness of
unions in influencing the political stance of their members.
Research on citizens’ preferences regarding trade policy may provide some insight on
this question. A number of earlier studies have found that union membership is, on average,
associated with lower support for free trade (Balistreri, 1997; Mayda and Rodrik, 2005; Scheve
and Slaughter, 2001b). When discussing this empirical association, authors often surmise
that it may be the outcome of unions’ ongoing communications on the matter with their
membership (e.g. Mansfield and Mutz (2009: 431,436)). Yet, other than conjecture, those
studies provide no evidence regarding the intensity of this communication of information, or
that this communication has any causal impact on shaping members’ attitudes.
Cognizant of this deficiency, Ahlquist, Clayton, and Levi (2014) provide what is arguably
7
the most careful and nuanced new set of insights on the matter. Focusing on a case study
of a dockworkers’ union (The International Longshore and Warehouse Union, or ILWU) and
using a survey of workers in three localities (Los Angeles, Seattle and Tacoma), the authors
employ a matching procedure to get an estimate of the “union effect”. Overall, they find
that members of the ILWU were more willing than non-members to support a protectionist
stance on trade, even though trade openness was highly beneficial to their own employment.
The authors propose that this seemingly puzzling result is evidence that the union was
able to influence its members to adopt a solidaristic stance, i.e. to oppose a policy that
was injurious to the broader class of workers. The study combines survey data and a rich
historical account of the ILWUs position on trade policy over a period of six decades. Yet it
remains unclear whether the findings from this specific case can be generalized to explain the
impact of unions on workers’ policy preferences in the broader economy. Overall, does union
membership instill a class-based approach to trade policy, thus having a largely homogenizing
effect on workers’ preferences, or do the preferences of unionized workers vary systematically
in a way that reflects the divergent interests of each industry?
To address these questions, one must not only investigate the impact of unions on a
broader set of sectors, but also explore the mechanism underlying union influence.3 If unions
are a source of crystallizing interests and shaping preferences of members, systematic evidence
should show that members: (i) are aware of the information provided by their unions; (ii)
correctly interpret the union’s stance on the matter; and (iii) adopt the position touted by
the unions. While these outcomes are at least plausible ex ante, convincing empirical research
on all three questions is lacking. In the next sections of this paper we aim to provide new
insights that address each of these contentions in turn.
3
Ahlquist and Levi (2013) provide evidence on the importance of the educational mechanism, showing
how the ILWU informs and shaping its members’ views. They also compare this to the approach taken by
to two transport unions and the Waterside Workers Federation in Australia. Our investigatation allows us
to assess how prevalent this educational role is in other unions’ relationship with their members in different
sectors throughout the economy.
8
Table 1: Descriptive Statistics for Selected Industries
Industry:
Manufacturing
Food products
Chemical
Transportation equipment
Computer electronics
Fabricated metal products
Services
Data processing and internet
Financial
Telecommunications
Construction
Nursing and residential care
Ambulatory health care
Education
Total
Output
Trade Share Median Sample
Employed per Worker Balance
BA Income
Size
(1000s)
($)
(B$) Degree
($)
(#)
1,485
850
1,607
1,248
1,528
292,093
8,400
546,482
-3,100
362,878 -14,000
412,519 -110,000
163,973
-9,900
22%
40%
24%
48%
15%
51,000
88,945
76,005
96,004
61,570
218
225
270
349
352
395
858
1,022
7,215
3,008
5,661
3,037
359,059
507,517
470,191
119,281
43,584
112,263
51,309
45% 82,557
65% 110,067
34% 83,000
15% 55,197
18%
4,590
48% 73,067
65% 79,235
320
375
375
393
382
446
607
0
41
2
0
0
0
13
Source: March Supplement of Current Population Survey 2009;
2010 United States International Trade Commission data on imports and exports
3
Data and Empirical Strategy
Our analysis uses novel survey data of more than 4,000 American workers employed in a set
of selected industries. The survey design followed a customized two-stage sampling approach.
First, a set of 12 key industries were identified based on several criteria reflecting variation in
their exposure to the impacts of globalization (e.g., factor intensity, value-added per worker,
trade balance, and exposure to offshoring activity).4 Then, from each of those targeted industries, a sizable number of currently employed workers were recruited by YouGov/Polimetrix
to participate in an online survey fielded between September 2010 and February 2011.5
To gain greater variation in the industries’ exposure to international commerce, the survey
included firms in both manufacturing and services. As Table 1 shows, the industries provide
4
Industries are classified at the 3 digit NAICS level.
The data was collected as part of the Harvard Globalization Survey in which Margalit was a co-PI. See
Hainmueller, Hiscox, and Margalit (2013) for a more detailed description of the survey. We thank the other
PIs for allowing us to use the data for this study.
5
9
wide variation of values along a set of pertinent dimensions. For example, as measured by
the value added per worker, the sample includes highly skill-intensive industries (e.g., chemical manufacturing and financial services), industries with a mediocre level of skill-intensity
(e.g., transportation equipment and computer electronics manufacturing), and industries
with very low skill intensity (e.g., construction and nursing). With respect to trade balance,
the selected industries include import-competing industries (e.g., transportation equipment
and computer electronics manufacturing), non-tradables (e.g., health services) and exportoriented industries (e.g., food manufacturing).
The data include responses from 497 union members, which represent about 12% of
the sample. However, union membership rates differ quite substantially across industries,
ranging from less than one percent in the financial services sector to over 35% in educational
services. As Figure 1 shows, the union membership rate obtained in the sample corresponds
quite well with the actual rate of union membership.6
The tone and exact wording of questions on trade policy could have a sizable effect on
the answers that respondents provide (Hiscox, 2006). The survey therefore asks a series of
questions that could potentially tap into different aspects of workers’ views on trade policy.
The analysis we present below relies on responses to all three questions:
We would like to learn about your views on trade with other countries - by trade
we mean American businesses and individuals buying goods from other countries
or selling goods to other countries.
• Overall, do you think trade with other countries should be expanded, reduced, or kept at its current level?
• Do you think that trade with other countries is good or bad for you and
your family?
• Do you think that trade with other countries is good or bad for the United
States as a whole?
To explore the link between union membership and views on trade, the survey also contains
6
Data is from the Union Membership and Coverage Database available at www.unionstats.com.
10
Educational Services
Transportation Equipment Manufacturing
Telecommunications
Construction of Building
Food Manufacturing
Fabricated Metal Product Manufacturing
Chemical Manufacturing
Nursing and Residential Care
Ambulatory Health Care Services
Internet and Data Processing Services
Computer and Electronic Product Manufacturing
Financial Services
0
10
20
30
40
Union Membership Rate in the Survey Sample (%)
Actual Union Membership Rate (%)
Figure 1: Union Membership Rate, by Industry
questions regarding the intensity of communication initiated by the union on trade policy, as
well as a question pertaining to the degree of information that the union members possess
about the unions’ position on the issue (see Appendix for exact question wording).
In order to examine the correspondence between how members assess their union’s stance
on trade policy and the union’s actual position, we generated a new measure of each union’s
“revealed preference” with respect to trade policy. We did so using information on each
union’s lobbying activity and official announcements on legislation related to trade. Focusing
on union activities during the two years prior to our study (from July 2008 to June 2010),
we coded the positions the unions took on trade-related bills and used this data to place
the unions along a trade protectionist-liberalizer scale. In total, we coded the activities
of 15 labor unions that represent 75% of the unionized survey respondents who provided
information on their affiliated unions.7
7
Of the 497 surveyed union members, 343 respondents provided information on their union affiliation.
11
We examine each union’s position on major trade-related bills. We focus on all the bills on
which at least one of the unions under study carried out an official (i.e., registered) lobbying
effort and those bills that had potential application across industries.8 For every bill, we
code each union’s position along a seven-point scale that ranges from ‘strongly protectionist’
(+3) to ‘strongly pro-trade’ (-3). The coding is based on the position expressed by the union
(i.e., pro- or anti-liberalization) and the number of quarters it registered lobbying activity for
or against the bill.9 For example, for bills on which a union lobbied against liberalization for
five quarters or more, the union’s position is coded as ‘strongly protectionist’. If the lobbying
took place for a shorter period of only 1 to 4 quarters, we assign a ‘protectionist’ (+2) score;
we code a union as ‘weakly protectionist’ (+1) score if it did not lobby on the bill but had
expressed a protectionist stance on the issue in its official pronouncements. Conversely, we
assign scores between -1 and -3 using the same coding criteria when the union takes a proliberalization stance. Finally, a ‘neutral position’ (0) is assigned if the labor union did not
express any view on the issue or conduct any related lobbying activity.10
Table 2 summarizes each union’s score on the trade protectionism scale. Since we examine
The majority provided the exact names of their union. For those respondents who only provided the codes of
their local branches (e.g. local 101) or their firm names, we were able to infer their union affiliation based on
residence, industry, and firm name. For that, we used the records of collective bargaining agreements between
firms and unions available through the Office of Labor-Management Standards (http://www.dol.gov/olms/
regs/compliance/cba/) and the Center for Union Facts (http://www.unionfacts.com/). Among the
remaining unionized workers not represented through the 15 unions, 57% belong to other unions in the
education sector which presumably hold a similar stance on policy issues with the primary unions in the
sector that we do examine, namely the National Education Association and the American Federation of
Teachers; the other 33% belong to 32 different organizations for which have only one or two members in the
entire sample.
8
We obtained each union’s lobbying reports from the Lobbying Disclosure Act Database available at
http://www.senate.gov/legislative/Public_Disclosure/LDA_reports.htm. We then examined all the
issues classified as trade-related issues according to the general issue area code in the lobbying report. See
Appendix for a detailed description on the bills and the selection criteria.
9
Lobbying reports often do not contain information on the organization’s actual position on the lobbied
issue. We therefore use the union’s official letters to Congress or public statements to code whether the
union’s stance is pro-trade or protectionist (i.e., protectionist or liberalizing).
10
While we distinguish a strong stance (+3 or -3) from a relatively weaker stance (+2 or -2) using the
threshold of lobbying for five quarters, the measured union stance remains robust when we revise the threshold
to four or to six quarters instead.
12
13
1)
2)
3)
4)
5)
+1
0
+1
+2
0
0
+1
+2
+3
+3
+1
+3
+3
+3
+2
0
0
0
0
0
0
0
+2
+3
+3
+1
+3
+3
+2
0
-1
0
0
0
0
0
0
0
+3
+3
+1
+2
+3
+3
0
+1
0
0
+2
0
0
0
0
+1
+1
+1
0
+1
0
+2
0
0
0
0
0
0
0
0
+2
0
0
0
+1
0
0
0
0
+1
0
0
0
+1
+1
+3
0
+1
+1
+2
+1
0
+1
0
+2
0
0
0
0
0
+1
0
+1
+2
+2
+3
0
0
0
0
0
0
0
0
0
+1
0
0
+3
0
+3
0
0
0
0
0
0
0
0
+1
+2
+1
+1
+1
+1
0
+2
0
0
0
0
0
0
0
0
0
0
0
+2
0
0
0
+4
0
+6
+5
0
0
+3
+9
+22
+13
+8
+20
+19
+15
+8
Total
Import competing industries: fabricated metal manufacturing, transportation equipment manufacturing, chemical manufacturing.
Building construction; * Intl Brotherhood of Teamsters also encompasses educational service, ambulatory service, and food manufacturing sectors.
Telecommunication
Education services, nursing, and ambulatory health
Export oriented industry: food manufacturing
+2
0
5) UFCW: Food/Commercial
BCTGM: Bakery/Tobacco
+3
3) CWA: Communications
+2
+1
0
0
+1
+3
+2
+1
IBT: Teamsters
2) UBC: Carpenters
IBEW: Electrical Workers
AFGE: Government Employees
SEIU: Service Employees
4) NEA: Natl Education Assn
AFT: Federation of Teachers
AFSCME: St./Cty./Mun.
+3
+3
0
+2
USW: Steelworkers
IAM: Machinists
1)
UAW: Auto Workers
IFPTE: Technical Engineers
TRADE Colombia Panama Korea Peru NAFTA TAA Currency Enft BuyAmr Recip
Table 2: Union’s Position on Each Issue and Calculated Protectionism Score
each union’s position on eleven different trade bills, the score could theoretically range from
a low of -33 (pro-free trade) to a high of +33 (protectionist). As expected, unions operating
in the import competing sectors – fabricated metal, transportation equipment, and chemical
manufacturing – exhibit the most protectionist stance. We also find, unsurprisingly, that the
least protectionist unions operate in the export oriented sector, food manufacturing, and in
the non-tradable service sectors of education, nursing, and ambulatory health services. Somewhat surprisingly, unions in the building construction and the telecommunication sectors,
both of which are not significantly affected by the flows of international trade, nonetheless
take a relatively protectionist stance on the trade bills under study.
Since not all union members in the sample provided the specific name of the union to
which they belong, we also generate an average protectionism score for each industry. For
instance, among building construction workers who belong to one of the fifteen unions, 47%
are members of the United Brotherhood of Carpenters (protectionism score: +13), 41% are
members of the International Brotherhood of Electrical Workers (protectionism score: +8),
and the remaining 12% belong to the International Brotherhood of Teamsters (protectionism score: +22). We therefore generate the average protectionism score in proportion to the
share of members from each union and assign the building construction industry an average
score of +12.11 We conduct the same method for calculating the average protectionism score
for the other industries, as well.12 As expected, the import competing sectors are calculated
to have the highest protectionism score: chemical manufacturing industry has the highest
11
The calculation is: 0.47*13+0.41*8+0.12*22=12. To ensure the external validity of our weighting
scheme, we also check the share of members in each union by industry using information from the online listing of collective bargaining agreements between firms and unions available at http://www.dol.gov/
olms/regs/compliance/cba/. The Office of Labor-Management Standards provides 2,573 files of collective
bargaining agreements along with the number of workers and the NAICS industry classification of firms
involved in the agreement. The calculation using the collective bargaining agreement files yields a similar
industry average protectionism score as our original calculation. We choose to use the weighting scheme
based on our sample of workers instead of the one based on the collective bargaining agreement files because
there is no available agreement file for the sectors of ambulatory healthcare, nursing, and residential care.
12
We do not look at the sectors of data processing, securities, and computer and electronic manufacturing,
because the share of workers who belong to a labor union (<10) is too small to be meaningful for analysis.
14
score (+19.8), followed by fabricated metal manufacturing (+16.2) and transportation manufacturing industries (+14.6). The industry average score for food manufacturing sector is
also relatively high (+11.4) because almost half of the members belong to two protectionist
unions, namely the UAW and the International Brotherhood of Teamsters.
4
Results
4.1
Do Union Members Have Different Policy Preferences?
We begin by presenting an unconditional comparison of union members’ trade policy preferences and those of non-members working in the same industry. Based on responses to the
three survey questions on trade described earlier, the panels in Figure 2 present a comparison
of the share of respondents in each group who: i) support reducing trade levels, ii) have a
negative perception of trade’s impact on self, and iii) have a negative perception of trade’s
impact on the United States as a whole
13
.
Trade Levels (%)
Trade on Self (%)
Trade on US (%)
Transportation Equipment
Fabricated Metal
Chemical Manufacturing
Food Manufacturing
Telecommunication
Building Construction
Ambulatory Health Care
Education
Nursing/Residential Care
0
.2
.4
.6
0
.2
Union Member
.4
.6
0
.2
.4
Non-Member
Figure 2: Trade Policy Views of Union Members vs Non-Members, by Industry
13
The binary measure uses the bottom two categories on the five-point scale to classify the opposition to
trade.
15
.6
The graphs show that union members’ policy preferences are different from those of
non-members, but the impact of union membership is not uniform across industries. In
the manufacturing industries – transportation equipment, fabricated metal, chemical manufacturing, and food manufacturing – union members tend to hold more negative attitudes
toward free trade than non-members. Yet, the difference in view associated with union
membership is not homogeneous across the service industries. For example, union members
employed in building construction are more opposed to trade expansion than non-members,
but the opposite pattern is registered in nursing and residential care, as well as is in the
ambulatory health care industries, where unionized workers are less protectionist. Finally,
we find little difference between the preferences of union members and non-members in both
the telecommunication and education sectors.
To get a better sense of the overall ‘union effect’ across all industries, while taking account
of the main potential confounders, we conduct a nearest-neighbor matching exercise. In this
exercise we match each union member in our sample with a non-unionized worker who is
employed in the same industry and is also of the same gender, ethnicity, martial status and
education level as the union member.14 After the requirement for exact matching on these
five criteria is fulfilled, the matching algorithm is instructed to seek the closest observation
in terms of income level and age.15 With the matched data, we estimate a probit regression
model calculating the average treatment effect of union membership on all three dependent
variables.
The results, presented in Figure 3, indicate that the average ‘union effect’ is indeed
considerable: union members are about 4 percentage points more likely to support reduction
in levels of trade than similar workers employed in the same industry who do not belong to a
union and about 5 percentage points more likely to perceive that trade is adversely affecting
14
15
For education level, we use a binary indicator of completing a 4-year college degree.
We do not match on the respondent’s party identification that could be influenced by union membership.
16
oneself. The largest effect is registered with regards to the view that trade is harming the
U.S. as whole, where the estimated union effect is an increase of 9 percentage points. As
the figure shows, even taking account of the uncertainty in the estimates, the union effect
is statistically distinguishable from zero at the 95% for two of the dependent variables and
90% for the third measure.
Trade Levels
Trade on Self
Trade on US
0.00
0.05
0.10
0.15
Figure 3: Average Treatment Effects (ATEs) of Union Membership
The question, of course, is what accounts for this ‘union effect’ and for the variation in
its size and sign across industries (as seen in Figure 2). Conditional on the impact of trade
openness on different sectors of the economy, when should we expect differences between
the trade attitudes of union members and those of other workers employed in the same
industry to be most significant? Two plausible answers come to mind. Some labor unions
might be more strongly opposed to trade expansion and thus more active than other unions
in communicating their views to their members. Alternatively, the variation we observe
may perhaps be explained by different sorting effects. Since labor unions across industries
represent different interests, they may attract as members those individuals who to begin with
hold similar views to those of the unions (i.e. self-selection mechanism). In other words, the
first mechanism holds that unions shape the views of their members through communication
17
and information provision, while the second mechanism suggests that unions merely echo
the preferences of their members, not shape them. The following sections present a number
of empirical tests that evaluate the relative validity of these two lines of explanation.
4.2
Unions as Information Providers
To evaluate the validity of the ‘information provision’ mechanism, we begin by using descriptive data to examine the basic expectation that unions do in fact communicate with their
members on the issue of trade policy. We then explore the extent to which members are
familiar with their union’s stance on the issue.
Our first analysis examines the issues that unions discuss most prominently in their
communications with their members. As part of the survey, respondents who belong to unions
were asked to list up to three issues that were most frequently addressed in their union’s
communications. The answers to this open-ended question, presented in Figure 4, indicate
that in some industries a considerable share of members describe trade as one of the three
most discussed issues by the union: 58% of the respondents belonging to the United Auto
Workers, 25% from the International Association of Machinists, and 18% from the United
Steelworkers. In sharp contrast, none of the respondents from two of the least protectionist
unions – the American Federation of Teachers and the Service Employees International Union
– listed trade as a frequently discussed issue. This pattern is consistent with the notion
that labor unions serve as information providers on trade issues. Moreover, the positive
relationship we observe between the intensity of communication on trade issues and our
measure of unions’ protectionist stance suggests that the latter is valid.
Next, we explore the degree to which workers are in fact familiar with their union’s policy
stance on trade. The panels in Figure 5 present the share of members who: i) answered that
they had received at least three communications from their union in the past year on the
issue of trade; ii) are either somewhat, or very familiar with the union’s position on trade;
18
United Auto Workers
IAM: Machinists
United Steelworkers
Trade
Union Benefit
Wage
Wage
Immigration
Union Benefit
Union Benefit
Trade
Trade
Politics
Health Care
Politics
Contract
Wage
Job Security
Health Care
Politics
Immigration
Job Security
Job Security
Health Care
Immigration
Contract
Contract
0 .1 .2 .3 .4 .5 .6 .7
Communication Workers of America
Health Care
0 .1 .2 .3 .4 .5 .6 .7
American Federation of Teachers
0 .1 .2 .3 .4 .5 .6 .7
Service Employees International Union
Union Benefit
Wage
Job Security
Wage
Union Benefit
Wage
Health Care
Health Care
Contract
Job Security
Job Security
Politics
Politics
Politics
Union Benefit
Contract
Contract
Trade
Trade
Trade
Immigration
Immigration
Immigration
0
.1 .2 .3 .4 .5 .6 .7
0
.1 .2 .3 .4 .5 .6 .7
0
.1
.2
.3
.4
.5
.6
Figure 4: Issues Unions Discuss Most Frequently
Note: The unionized respondents were asked to list up to three issues that their union discussed most
frequently in its communications to the respondent. We reclassified open-ended responses to eight categories
presented in the figure, leaving out some answers that appear only rarely in the responses.
and iii) think that their union advocates reducing trade. The unions are sorted along the
vertical axis by their protectionism score which, as described earlier, is based on the union’s
lobbying activity and official statements.
The left panel indicates that members of more protectionist unions typically received
more communication from their organization on trade-related issues. Accordingly, those
members also tend to express greater familiarity with their union’s stance on the issue of
trade (center panel), and to describe their union as protectionist (right panel). In the case
of the more protectionist unions such as the United Auto Workers, the United Steelworkers,
19
.7
or the United Brotherhood of Carpenters, over 70% of the members correctly note that
their union supports reduction of trade levels. In contrast, in the American Federation of
Teachers – the least protectionist union in our sample – only a fraction describe the union as
protectionist, while the large majority thinks their union is either in favor of keeping trade
at its current level (62%) or expanding it (19%).
Teamsters
Union's Protectionism Score
Steelworkers
IAM:Machinists
United Auto Workers
Carpenters
Communication Workers
IFPTE
IBEW
Govt Employees
Service Employees
United Food/Commercial
State/County/Municipal
Natl Education Assn
Federation of Teachers
0
.1
.2
.3
.4
.5 0
Communication on Trade
Three Times or More in a Year (%)
.2
.4
.6
.8
Familiar with
Union's View on Trade (%)
1 0
.2
.4
.6
.8
1
Members' Perception of
Union as Protectionist (%)
Figure 5: Union Communications on Trade and Members’ Knowledge
The findings that Figure 5 highlights clearly lend support to the ‘information provision’
mechanism laid out above. Indeed, we find that members, especially those belonging to
protectionist unions, do receive correspondence on trade-related issues and learn about the
union’s policy stance. Yet to what extent does learning about the union’s policy position
have an impact on the members’ own attitudes? The next section explores this question in
some detail.
4.3
Do Members Internalize Information from the Union?
To assess whether communication from unions affects the preferences of workers on the policy,
we examine the alignment between the stance of the union and its members’ own attitudes
20
toward trade openness. The upper panels in Figure 6 compare the stance of each union with
the views of its members, presenting in each sub-graph the responses to one of the three
dependent variables questions. The graphs show quite vividly that members’ own attitudes
on the issue of trade are positively associated with the protectionism score of their union.
.6
Trade on US (%)
.3
.4
.2
.5
.1
.1
.1
Trade on Self (%)
.3
.4
.2
.5
Trade Levels (%)
.2
.3
.4
.5
.6
.6
Indeed, the positive association appears across all three measures of trade preferences.
.6
Trade on US (%)
.4
.3
.5
.1
0
5
10
15
20
Industry Average Protectionism Score
5
10
15
20
Union's Protectionism Score
.2
.6
Trade on Self (%)
.3
.5
.2
.4
.1
0
.6
5
10
15
20
Union's Protectionism Score
Trade Levels (%)
.3
.4
.5
0
.2
5
10
15
20
Union's Protectionism Score
.1
0
0
5
10
15
20
Industry Average Protectionism Score
Union Members
0
5
10
15
20
Industry Average Protectionism Score
Non-Members
Figure 6: Alignment between Union’s Stance and Workers’ Policy Preferences across Unions
This finding, however, is subject to an obvious concern about endogeneity and potential
spuriousness: The association might be simply driven by another factor that shapes both
21
the unions’ stance as well as that of its members. For example, workers and unions in
import competing sectors might be more protectionist than others simply because of the
adverse consequences that exposure to foreign competition poses to them. As a first way of
dealing with this possibility, we compare the trade preferences of union members with those
of non-union workers employed in the same industry. If the association is driven by some
industry-level characteristic, we should observe the same pattern within an industry among
both union and non-union members. Yet empirically, that does not appear to be the case.
The lower panels in Figure 6 present the share of union members and non-members
holding negative views toward international trade, and plot them against the industry’s
protectionism score. The graphs highlight that the average protectionism score of unions in
each industry is positively correlated with union members’ trade preferences, but not with
those of non-members. While members from an industry represented by more protectionist
unions appear to be more likely to hold negative views on trade openness, this relationship
does not hold among non-members. This suggests that workers from the same industry not
only differ in their views on trade as a function of whether or not they belong to a union,
but also that the differences reveal a distinct pattern: the former hold views that correspond
to those of the union while those of the latter group do not.
This striking pattern which Figure 6 highlights finds support also when tested formally.
Specifically, we estimate a probit model:
P robit(Yi ) =α + β1 Industry Protectionism Scorei + β2 Union Memberi
+ β3 Industry Protectionism Stance*Union Memberi + θControlsi + i ,
where Yi is a binary measure of individual i’s view on trade and Average Protectionism
Stance is the average protectionism score for unions in i’s industry of employment. Union
Member is a binary indicator for union members, and Industry Protectionism Stance*Union
Member is the key variable of interest, capturing the interaction between the two variables.
22
The specification also controls for education, age, income, gender, race and martial status.
Table 3: Union Average Protectionism Stance and Worker’s View on Trade (Members vs.
Non-Members)
(1)
(2)
(3)
Trade Level
Observations
(5)
(6)
Trade on Self
(7)
(8)
(9)
Trade on US
-0.004∗
(0.002)
-0.063+
(0.035)
0.011∗∗
(0.004)
-0.002
(0.002)
-0.018
(0.039)
0.007+
(0.004)
0.001
(0.001)
0.038+
(0.021)
-0.000
(0.001)
-0.015
(0.032)
0.007∗
(0.003)
0.001
(0.002)
0.020
(0.035)
0.003
(0.003)
-0.000
(0.001)
0.057∗
(0.023)
-0.002
(0.002)
-0.023
(0.033)
0.010∗∗
(0.003)
-0.000
(0.002)
0.015
(0.037)
0.007+
(0.004)
No
No
Yes
No
No
Yes
No
No
Yes
3194
3194
2911
3194
3194
2911
3194
3194
2911
Industry Protectionism Score -0.002
(0.001)
Union Member
0.024
(0.024)
Industry Protectionism Score
*Union Member
Demographic Controls
(4)
Marginal effects; Standard errors in parentheses
+
p < 0.10, ∗ p < 0.05, ∗∗ p < 0.01
The results of the estimation are presented as marginal effects in Table 3. The results
suggest that the unions’ stance on trade has a notable impact on the attitudes of the union
members, but not on the attitudes of the non-members. We find a positive and statistically
significant coefficient for the interaction term β3 , but the coefficient for β1 is in most specifications very small in magnitude and statistically indistinguishable from zero. This indicates
that union members exhibit attitudes on trade that much more closely resemble the position
staked by the unions than the attitudes of the non-members employed in the same industry.
5
Treatment versus Selection
The results presented so far are consistent with the notion that, among other functions,
unions are information providers that exert effective influence on their members’ policy
preferences. Yet, as noted above, these findings may also reflect a self-selection process:
If workers’ decision to join a union takes into account the position of the union on traderelated issues, and if those who join protectionist unions are also more engaged in reading
the communications they receive on the topic, the results in Figure 6 could be an outcome
23
of a reverse causal process. To address this possibility and test whether a selection effect
accounts for the results, this section presents a number of inferential tests designed to help
tease out between the competing explanations.
5.1
Cross-State Legal Differences and the Union Effect
To address the possibility that self-selection accounts for the ‘union effect’ detected in the
analyses above, we leverage state-level differences in their “Right-to-Work” laws, which refer
to the legal conditions that govern the ability of workers to join or opt out of a union.
Since the introduction of the Right-to-Work (henceforth RTW) provision in 1947, as part
of the Taft-Hartley Act, individual states in the U.S. have had the option of enacting a law
that prohibits union “security agreements”. This means that in states that adopt the RTW
law, labor unions cannot legally require workers to pay union dues. The implication is that
union membership in RTW states depends much more on individual workers’ own discretion
and is less a function of an institutional requirement to do so.16 Indeed, union membership
rates are systematically much higher in non-RTW states, even within the same industry
(see Figure 7). This difference in the regulation of union membership across states allows
us to test the self-selection explanation in the following manner: if self-selection accounts
for members’ preferences, the effect of union membership should be larger in those states in
which membership is much more likely to arise from a worker’s choice.
To test this proposition, we estimate the following binary probit model:
P robit(Yi ) = α + β1 Unioni + β2 RTWi + β3 Union ∗ RTWi + γIndustryi + θControlsi + i ,
where Yi is a binary measure of respondents’ attitudes toward international trade. Union is
a binary indicator for an individual i’s union membership, and RTW is a binary variable
16
For an overview of the Right-to-Work law, see Collins (2012).
24
Educational Services
Transportation Equipment Manufacturing
Telecommunications
Construction of Building
Food Manufacturing
Fabricated Metal Product Manufacturing
Chemical Manufacturing
Nursing and Residential Care
Ambulatory Health Care Services
Computer and Electronic Product Manufacturing
Internet and Data Processing Services
Financial Services
0
10
20
30
40
Union Membership (%)
Right-to-Work States
50
Non Right-to-Work States
Figure 7: Unionization Rate by Industry and Right-to-Work Law Status
that takes the value 1 if i resides in a state that had adopted the ‘Right to Work’ law at the
time of the survey. The key parameter of interest is the coefficient β3 on the interaction term
Union * RTW. If we find that the interaction term is sizable and significant, that would point
strongly toward a selection-based explanation, as it would indicate that the ‘union effect’ –
the difference in attitudes of union members and non-members in their industry – are less
pronounced when workers are “pushed” into their union membership status.
The model also includes fixed effects for Industry as well as Controls, a vector of individual characteristics (income, gender, race, age, education, and martial status).17 In some
models we also include a measure of party identification (an ordinal scale ranging from strong
democrat (1) to strong republican (7)). In the last column of each set of specifications, we
also include the control variables interacted with RTW, to control for the possibility that
17
Income is a 14-level ordinal variable, ranging from less than $10,000 (1) to more than $150,000 (14);
education is 6-level ordinal variable, ranging from no high school education (1) to a graduate degree (6).
25
individual characteristics may also have varying effects under the different legal settings.
We expect labor unions to affect the policy preferences of their members when they
actively disseminate policy-related information to their membership. Thus, to provide an
effective test of the treatment mechanism, we conduct a split-sample analysis in which we
estimate the model separately for industries in which the average protectionism score is high
– transportation equipment, chemical, fabricated metal, and food manufacturing, telecommunication and building construction industries – and the rest of the sample, i.e. industries
that score low on the protectionism score.18
We present the estimation results with all three dependent variables in Table 4 – the
results for industries with high average protectionism score in the top panel of the table, and
the results for the rest of the industries in the lower panel. For each dependent variable,
we begin by including only a set of basic covariates. The coefficient on Union Member is
positive and statistically significant in the industry groups represented by the protectionist
unions (top panel), yet it is not significant in the less protectionist industries (lower panel).
This result, which holds across all the model specifications, suggests that the effect of union
membership is conditional on the firmness of the union’s stance on the policy issue in question, perhaps because those unions communicate their stance on the issue more intensely to
their membership.
Yet as explained above, the key coefficient of interest is the interaction term Union
Member * RTW. Notably, this interaction term is not statistically significant in any of the
specifications in the protectionist industries: this is case when we include only the basic set
of covariates, when we control also for respondents’ partisan affiliation, as well as when we
interact RTW with all other covariates. Moreover, the substantive effect of the interaction
term is either very close to zero or small and slightly negative, a finding that is inconsistent
18
None of the labor unions in the sample is actively advancing a pro-trade stance. The difference is thus
between unions strongly opposed to trade-liberalizing bills and unions that are not strongly opposed.
26
27
(2)
(3)
Trade Levels
(4)
(5)
(6)
(7)
Trade on Self
(8)
(9)
1634
Observations
1634
No
No
(2)
(3)
Trade Levels
1607
Yes
No
2196
Observations
2196
No
No
2160
Yes
No
(5)
1634
No
No
1607
Yes
No
(6)
(7)
Trade on Self
1634
No
No
(8)
1607
Yes
Yes
(9)
1634
No
No
1607
Yes
No
0.146∗∗
(0.042)
0.024
(0.024)
-0.066
(0.059)
(10)
(11)
Trade on US
1634
No
No
0.145∗∗
(0.042)
0.016
(0.023)
-0.049
(0.062)
(10)
(11)
Trade on US
(12)
1607
Yes
Yes
0.146∗∗
(0.042)
0.040
(0.170)
-0.063
(0.060)
(12)
2160
Yes
Yes
2196
No
No
2196
No
No
2160
Yes
No
2160
Yes
Yes
2196
No
No
2196
No
No
2160
Yes
No
2160
Yes
Yes
0.010
0.012
0.024
0.019
0.011
0.010
0.008
0.002 -0.002
(0.038) (0.028) (0.032) (0.032) (0.032) (0.030) (0.033) (0.033) (0.033)
0.051 -0.004 0.000
0.000
-0.054 -0.017 -0.018 -0.017 0.054
(0.152) (0.017) (0.018) (0.018) (0.116) (0.018) (0.019) (0.019) (0.139)
-0.122∗
-0.049 -0.047 -0.028
0.008
0.011
0.025
(0.055)
(0.050) (0.051) (0.057)
(0.066) (0.067) (0.070)
(4)
1607
Yes
Yes
Marginal effects; Standard errors in parentheses
+
p < 0.10, ∗ p < 0.05, ∗∗ p < 0.01
All models include fixed effects for Industry as well as Controls (income, gender, race, age, education, and martial status.
Full Interactions include the demographic controls interacted with RTW.
No
No
Party ID Control
Full Interactions
-0.013 0.018
0.017
(0.033) (0.038) (0.038)
RTW
-0.004 0.007
0.007
(0.020) (0.021) (0.021)
RTW*Union Member
-0.129∗ -0.130∗
(0.054) (0.053)
Union Member
(1)
Not Strongly Protectionist Unions
No
No
Party ID Control
Full Interactions
Union Member
0.097∗∗ 0.090∗ 0.096∗ 0.097∗ 0.070∗ 0.084∗ 0.084∗ 0.094∗ 0.130∗∗
(0.037) (0.042) (0.043) (0.044) (0.033) (0.038) (0.039) (0.040) (0.036)
RTW
0.021
0.018
0.023 -0.076 -0.002 0.003
0.009
0.260
0.011
(0.024) (0.025) (0.026) (0.174) (0.021) (0.022) (0.022) (0.170) (0.022)
RTW*Union Member
0.026
0.007
0.008
-0.053 -0.069 -0.086+
(0.082) (0.081) (0.083)
(0.058) (0.054) (0.050)
(1)
Strongly Protectionist Unions
Table 4: Effect of Union Membership on Attitudes toward Trade
with the selection mechanism being at play. In the less protectionist industries, the finding is
similar. The interaction coefficient is either substantively close to zero or negatively signed.
In sum then, the results clearly go against the prediction that arises if a selection mechanism
accounts for the distinct trade policy preferences of union members.
To assess the substantive effect of union membership on trade attitudes, we estimate the
probability that a worker with characteristics of the sample median supports a reduction in
trade levels.19 A non-union member with such characteristics is, on average, 21% likely to
support a more protectionist measure, but union membership increases this probability by
8 percentage points to over 29%. This represents about a 41% increase over the baseline
level. We also examine the effect of union membership in the less protectionist group of
industries.20 In this case, union membership increases the predicted probability of supporting
trade reduction by less than 2 percentage points on average, yet the point estimate is well
below conventional levels of statistical significance.
The substantive effect of union membership on the probability that a worker perceives
a negative impact of trade on oneself and family is even more considerable. A worker with
characteristics of the sample median has a 17% likelihood of perceiving trade as harmful, but
this estimate increases by 8 percentage points among union members with similar characteristics, representing a 50% increase compared to the baseline estimate. Note that this effect is
comparable in size to that associated with education, a variable that is widely documented
as an important determinant of trade preferences (e.g., Hainmueller and Hiscox (2006)).21
19
We estimate the predicted probability based on the second model and set age at its mean value and all
other categorical variables at their median values, assuming a white male, married, with 4 years of college
education. The industry is set to transportation equipment industry when estimating the model for more
protectionist group of industries.
20
We estimate the probability setting the industry category to ambulatory health care industry.
21
A change from a high school diploma to a 4-year college graduate is associated with a decrease in the
predicted probability of a negative perception of trade by 10 percentage points.
28
5.2
Members’ Preferences when the Union Changes Position
If unions affect the policy positions of their members by providing policy relevant information,
we would expect that following a change in the policy stance of the union, a corresponding
change in the view of their members would also take place. In contrast, we would not expect
this to occur if members join the union because of their affiliation with its (original) stance
on trade. In this section we examine the effect of exactly such type of reversal in a union
stance: the sudden and fairly dramatic shift in the United Auto Workers (UAW) position
toward a major trade liberalization deal. This examination thus provides a second setting
for testing the competing explanations for the previously observed union effect on members’
trade preferences.
For many years, the UAW, a labor union representing workers primarily in the auto
industry, had been consistently and strongly opposed to the expansion of U.S. trade. It was
also part of a vocal opposition to the signing of trade agreements with Colombia and with
Korea, agreements that were debated exactly around the time of the survey. With respect to
the latter agreement, the UAW’s official statement from April 2010 summarized its position
as follows: “The UAW strongly opposes the free-trade deal negotiated by President Bush with
South Korea (KORUS FTA) in April 2007, and has reiterated that opposition to the Obama
administration and to Congress. The poorly negotiated and misguided auto provisions of
the KORUS FTA would further open the U.S. market to increased automotive imports from
Korea...” The statement ended by calling the union members to “Tell Congress that this
free-trade deal would lead to a surge in automotive imports from South Korea, worsening
our lopsided auto trade deficit and threatening the jobs of tens of thousands of American
workers.”22
Yet, following intense lobbying and negotiations with the Obama administration, a set of
22
The full statement is at http://www.uaw.org/page/international-trade-and-investment-policy.
29
changes advocated by the union were incorporated into the revised agreement. On December
6th of that same year, the union made a statement pronouncing that “the changes announced
to the U.S.-Korea Free Trade Agreement today are a dramatic step toward changing from
a one-way street to a two-way street for trade between the U.S. and South Korea. These
changes represent an important opportunity to break open the Korean market for U.S. businesses and workers and boost American manufacturing jobs, particularly in the automotive
sector.” The announcement went on to detail the advantages of the revised trade deal: “We
believe an agreement was achieved that will protect current American auto jobs, that will
grow more American auto jobs, that includes labor and environmental commitments, and
that has important enforcement mechanisms.”23
How did this shift in the union’s position influence the views of the autoworkers on
trade? We examine the impact of the UAW’s pro-trade message by focusing on our sample
of auto industry workers. The survey includes 102 respondents from the auto industry, a
quarter of which participated in the survey after the UAW announced its support for the
free trade agreement.24 Using this sample, we compare the views of union members with
those of non-members before and after the UAW’s endorsement of the free trade agreement.
Figure 8 clearly demonstrates that union members working in the auto industry were significantly more protectionist than non-members before the shift, yet the level of support
for trade restrictions significantly decreased after the UAW endorsed the free trade agreement. Crucially, this change in attitudes toward trade liberalization is not observed among
non-members working in the same auto industry.
This pattern is very much consistent with the educational, information provision mechanism we put forth here that union members became more supportive of trade as they received
a pro-trade message from the union. Yet this observed pattern could potentially be explained
23
The full statement is at http://www.uaw.org/category/tags/korus.
The survey includes 270 respondents working in the transportation equipment industry, which encompasses aircraft manufacturing and aerospace manufacturing industries as well as automobile industry.
24
30
5
4
3
2
1
Mean (Support for Reducing Trade Levels)
Pre-Shift
Post-Shift
Pre-Shift
Non-Members
Post-Shift
Union Members
95% Confidence Interval
Figure 8: Support for Reducing Trade by Union Membership before and after the UAW Shift
also by reversed causality, namely that a shift in members’ trade preferences (following the
renegotiation of the trade deal) was itself the trigger for the change in the union’s public
stance. Such an explanation, however, is highly implausible given the very complicated and
technical nature of the changes made to the trade agreement. These included new provisions
on the schedule of tariff reductions, changes to the list of safety regulations, incorporation
of certain environmental standards, and the introduction of safeguard provisions pertaining
to Korean exports.25 The average worker might not even be aware that such changes were
made to the agreement, let along be able grasp how these technicalities would affect her
well-being without the assistance of the union’s message communicating and simplifying the
bottom line of these changes: openning markets to U.S. vehicles and increasing the number
American auto jobs.
Another alternative explanation for the shift that occurred only in union members’ prefer25
For instance, Korea agreed that motor vehicles produced by the U.S. manufacturer that sold no more
than 25,000 in Korea shall be deemed to comply with Korean safety standards if the motor vehicle is certified
to comply with the U.S. safety standards. see Appendix for detailed information on the revised agreement.
31
ences could be news consumption. If union members follow the news more than non-members,
perhaps the explanation is that members became supportive because of greater exposure to
the news coverage of the agreement. Against this possibility, it should first be noted that
the revised agreement was in fact severely criticized by key media outlets, some of which
took a stance that directly contradicted the union’s assessment of the deal.26 Thus, it is
unclear in what direction news consumption would influence one’s attitudes toward trade
before and after the change in the agreement. Notwithstanding, in the following estimation
we also control for respondents’ level of news consumption along with other potential confounding factors to ensure that the difference between union members and non-members are
not driven by other characteristics. We estimate the model:
P robit(Yi ) = α + β1 Unioni + β2 Post-Shifti + β3 Union ∗ Post-Shifti + θControlsi + i .
This specification is similar to the one estimated in the previous section, only here we include
a Post-Shift indicator instead of a binary variable denoting an RTW state. The Post-Shift
indicator variable takes the value 1 if individual i was interviewed after the UAW announced
its support for the KORUS FTA and the value 0 if interviewed before. In some models we also
include separate indicators for Michigan and Ohio, the two states in which the auto industry
is concentrated, as well as their interaction terms with a binary indicator for post-shift survey.
These controls are necessary to ensure that a finding is not driven simply by state specific
characteristics. The main interest in this analysis is of course the effect associated with
Union membership and the interaction term Union*Post-shift. We expect union members
interviewed before December 2010 to exhibit more intense protectionist attitudes than nonmembers because the former were exposed to the union’s message opposing the free trade
26
For example, The New York Times published an article under the headline, “Few New Jobs Expected
Soon from Free-Trade Agreement with South Korea,” which argued that “the effect of the agreement on
aggregate output and employment in the United States would likely be negligible.” (December 7, 2010).
32
deal. In addition, we expect that union members interviewed after the shift – and who
presumably were exposed to the pro-trade message from the union – to be less protectionist.
Table 5: Change in the Union’s Policy Position and Members’ Preferences
(1)
(2)
(3)
Trade Level
(4)
0.547∗∗ 0.497∗∗ 0.503∗∗ 0.478∗∗
(0.120) (0.140) (0.143) (0.156)
Post-Shift
0.062
0.021
-0.002
-0.042
(0.128) (0.136) (0.137) (0.171)
Post-Shift*Union Member -0.442∗∗ -0.465∗∗ -0.469∗∗ -0.501∗∗
(0.076) (0.076) (0.079) (0.072)
Union Member
(5)
0.401∗∗
(0.140)
-0.054
(0.120)
-0.067
(0.206)
(6)
(7)
Trade on Self
(8)
0.303+ 0.414∗ 0.374∗
(0.156) (0.164) (0.177)
-0.092 -0.119 -0.014
(0.122) (0.123) (0.157)
-0.022 0.046
0.155
(0.235) (0.258) (0.326)
Demographic Controls
News Consumption
Party ID
Auto States
No
No
No
No
Yes
No
No
No
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
No
No
No
No
Yes
No
No
No
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
Observations
102
98
97
97
102
98
97
97
Marginal effects; Standard errors in parentheses
+
p < 0.10, ∗ p < 0.05, ∗∗ p < 0.01
Demographic Controls: income, gender, race, age, education & martial status
News Consumption: A binary indicator taking a value of 1 if the respondent read a newspaper once a day or more.
Auto States: Michigan, Post-Shift*Michigan, Ohio, Post-Shift*Ohio
The estimation results are presented in Table 5. The results are indeed in line with the
expectations: The coefficient on Union Member is positive and statistically significant at the
0.01 level or higher in all specifications estimating support for trade reduction. In addition,
the coefficient on Union Member*Post-Shift is negative and highly significant in all models.
This is the case even when we control for respondent’s news consumption and ideological
affinity. Given that this result is estimated with a sub-sample of 102 auto workers, the consistency of the finding, both unconditionally and when controlling for a host of confounding
factors, is quite striking.
Turning to the right panel of the table in which we analyze respondents’ view of trade as
harmful to themselves and their families, we find that union membership was again associated
with a sizable and significant effect on the perception of trade as adversely affecting oneself.
33
However, in this case we observe a much weaker change following the union’s u-turn in
the pro-trade direction. Taken together, these results suggest that the union was able to
effectively influence the policy stance of its members after it publicly changed its position on
the signing of the KORUS agreement, but this shift did not reverse the members’ perception
that trade has overall been harmful to them and their families.
5.3
Bounding the “Union Effect”
The analyses presented herein demonstrate that the effect associated with union membership on the policy preferences of its members cannot be accounted for by a selection (i.e.
self-sorting) mechanism. Nonetheless, the evidence does not eliminate the possibility that
selection accounts for at least some, if not most of the so-called ‘union effect’. Thus, in the
final analysis we estimate a lower bound of the treatment in a condition where selection on
unobsevables may account for some of the estimated effect.
We do so using the method pioneered by Altonji, Elder, and Taber (2005) and recently
developed in Oster (2014), which estimates the bounds of a treatment effect based on coefficient movements after inclusion of controls. The logic is of this approach is straightforward:
If we assume that selection on observables is proportional to selection on unobservables, we
can examine how much coefficients change with the inclusion of observables and form an understanding of the sensitivity of a coefficient on unobservables. If the coefficient moves little
after the inclusion of controls, this suggests that the coefficient is robust to unobservables.
Yet, this movement must be scaled by movements in the R-squared because an uninformative control does not change the coefficient in a significant manner but also adds little to the
R-squared.27
Following these assumptions, we estimate linear regressions for all three dependent vari27
See appendix for a detailed explanation of the method in the context of this study.
34
ables.28 In the first stage we estimate the model when only including the indicator of union
membership; in the second stage we estimate the model with the full set of controls.29 We
then use the R-squared values to calculate the identified set of union treatment effects. As
before, we estimate the models separately for workers in industries represented by highly
protectionist unions and for workers in less protectionist industries.
Table 6: Identification of Lower Bound of Treatment Effect
Baseline Effect
(S.E.) [R2 ]
Controlled Effect
(S.E.) [R2 ]
Identified Set
Strongly Protectionist Unions
Trade Level
0.103 (.032) [.005] 0.087 (.034) [.087] [0.069, 0.087]
Trade on Self 0.082 (.029) [.004] 0.065 (.031) [.083] [0.047, 0.065]
Trade on US 0.127 (.030) [.010] 0.122 (.032) [.073] [0.117, 0.122]
Not Strongly Protectionist Unions
Trade Level
0.004 (.029) [.000] -0.020 (.032) [.113] [-0.045, -0.020]
Trade on Self 0.023 (.025) [.000] 0.001 (.028) [.073] [-0.021, 0.001]
Trade on US 0.033 (.027) [.001] 0.006 (.030) [.097] [-0.020, 0.006]
Table 6 summarizes the results. The identified set shows the lower and upper bounds
of the union treatment effect. The lower bound refers to the treatment effect when we
assume that the unobservables are as important as the observables in explaining the impact
of union membership on trade attitudes. The upper bound denotes the union treatment
effect when we assume that there is no selection on unobservables. Among protectionist
unions, the results show that the lower bound of the union effect is both positive and sizable
for all three dependent variables. For example, in the case of support for reduction in trade
levels, the lower bound is 0.069, which means that union members are about 7 percentage
points more likely to support a reduction of trade than non-members, even when we assume
that the unobservables are as important as the observables. This represents about a 35%
28
The estimation procedure requires a linear regression model. Similar to Altonji, Elder, and Taber (2005:
176), we estimate linear regression models by assuming that the unobservable selection bias in a probit model
is close to the bias in an OLS model.
29
Controls include income, gender, race, age education, martial status, and the fixed effects for industry
and state.
35
increase from the baseline rate. Crucially, in all three dependent variables, even the lower
bound of the estimated union effect is sizable, representing at least 76% of the upper bound
estimate. This indicates that the large bulk of the union effect is not due to selection on
unobservables. Finally, in the bottom three rows, which show the estimation results for the
less protectionist unions, the finding is very different: the union effect is either not robust to
selection on unobservables or substantively very close to zero.
6
Conclusion
Labor unions are conventionally seen as organizations fighting for better rights, wages and
benefits for workers. Despite wide agreement among scholars that these objectives guide
much of unions’ activity, the methods used by unions to obtain these objectives and the
consequences of their deployment are often deeply contested. This paper examines one route
by which labor unions pursue their objectives, namely through emboldening support for the
organization’s stance by communicating information to the union’s membership. To date,
little systematic information exists about the degree to which unions communicate with
their members on specific policy issues, and even less is known about the impact of these
communications on members’ preferences.
Exploring these issues, this paper has shown that unions do indeed communicate policyrelevant information to their membership on a regular basis, but that the frequency and
nature of these communications varies significantly across unions. Focusing on the issue of
trade policy, we find that unions that engage in substantial lobbying on trade-related bills
also tend to communicate regularly on the matter with their members. In fact, in some cases,
communication of information on trade policy eclipses correspondence even on traditional
union issues, such as wages or health care. Beyond documenting these patterns of communication, we provide evidence that those unions that engage in substantial communication
of information to their members are also able to influence the members’ attitudes toward
36
the union-held position. Thus, it appears that unions are not merely a “voice” of workers’
preferences, but also an effective institution that is able to systematically shape and cohere
that voice toward a given policy objective.
The rise in levels of international trade in recent decades poses a major challenge to
labor unions, as competition from overseas constrains their ability to negotiate improved
compensation or employment conditions for workers. Indeed, the weakening of unions in
the face of deepening globalization has been cited as a contributing factor to the rise in
wage inequality in various countries. Our results highlight a largely overlooked channel that
may account for this development, namely the ability of unions to influence their members’
political attitudes. Our findings suggest that unions may in fact be influencing the dispersion
of incomes not only directly, i.e. by negotiating improved compensation for workers in
the lower income rungs, but also indirectly by affecting the degree of public support for
inequality-reducing policies.
One might ask whether workers cannot discern their policy interests without the unions’
guidance on such policies. For sure, some workers can. Yet our analysis shows that elite
communication, in this case by unions, significantly affects the policy views of their members and accounts for variation in policy preferences even among similarly-skilled workers
employed in the same industry. These findings, together with results obtained from in depth
case studies (Ahlquist and Levi, 2013), suggest that for explaining individual attitudes on
a policy one must look beyond its distributive effects and also pay attention to citizens’
sources of information. As this paper demonstrates, these sources can have a rather strong
intervening effect.
Our findings pertain to the impact of unions in the U.S., with relatively low rates of
membership. One obvious extension of this study would be to examine our research question
but in countries with much higher union density rates, such as Denmark, Norway, or even
Canada. We conjecture that the union effect we would find in those countries would be even
37
larger than the one we identify in this study, since the strength of the unions outside the U.S.
allows them to invest more in communicating and educating their members. Whether that
is the case is of course an empirical question that hopefully future research would illuminate.
In prior research on public opinion, any consideration of a “union effect” on attitudes has
almost exclusively relied on the inclusion of an indicator variable denoting whether or not the
respondent belongs to a union. This approach assumes a homogeneous effect across unions.
Yet our study, which utilizes information not only on membership but also on the specific
unions the respondents’ belong to, highlights the significant variation both in the position
that unions take on the same issue as well as in the intensity in which they correspond
with their members about the issue. Thus, by estimating only the average union effect, as
most prior research has done, scholars have underestimated the impact of the more active
unions on the preferences of their members. This suggests that for addressing some questions
about the political consequences of unions, particularly those that seek to go beyond their
overall effect on the electorate, collecting information not just on membership but also on
the specific union affiliation could be very useful.
In recent years, perhaps due to the declining rates of union membership, the focus in
much of the research has shifted to exploring the influence of other institutions such as the
church or business lobbies on various political and electoral outcomes (Green, 2007; Smith
and Walker, 2013; Baumgartner et al., 2009). Yet even today, few organizations have the
broad reach and regular access to such sizable portions of the electorate as labor unions
do. Indeed, as the findings of this paper indicate, a meaningful understanding of the forces
shaping public preferences in today’s political environment requires taking a serious account
of unions’ significant impact.
38
References
Ahlquist, John, Amanda B. Clayton, and Margaret Levi. 2014. “Provoking Preferences:
Unionization, Workers’ Attitudes toward International Trade, and the ILWU puzzle.” International Organization 68 (1): 33–75.
Ahlquist, John, and Margaret Levi. 2013. In the Interests of Others: Leaders, Governance,
and Political Activism in Membership Organizations. Princeton: Princeton University
Press.
Alesina, Alberto, and Eliana La Ferrara. 2005. “Preferences for Redistribution in the Land
of Opportunities.” Journal of Public Economics 89 (5): 897–931.
Altonji, Joseph G, Todd E Elder, and Christopher R Taber. 2005. “Selection on Observed
and Unobserved Variables: Assessing the Effectiveness of Catholic Schools.” Journal of
Political Economy 113 (1): 151–184.
Asher, Herbert B, Eric Heberlig, Randall Ripley, and Karen Snyder. 2001. American Labor
Unions in the Electoral Arena. Rowman & Littlefield.
Baldwin, Robert E. 2003. “Decline of US Labor Unions and the Role of Trade, The.” Peterson
Institute Press: All Books .
Balistreri, Edward J. 1997. “The Performance of the Heckscher-Ohlin-Vanek Model in Predicting Endogenous Policy Forces at the Individual Level.” Canadian Journal of Economics
pp. 1–17.
Bartels, Larry M. 2009. Unequal democracy: The political economy of the new gilded age.
Princeton University Press.
Baumgartner, Frank R, Jeffrey M Berry, Marie Hojnacki, Beth L Leech, and David C Kimball. 2009. Lobbying and Policy Change: Who Wins, Who Loses, and Why. University of
Chicago Press.
Becher, Michael, and Daniel Stegmueller. N.d. “Rational Mobilization of Ideological Group
Members.” . Forthcoming.
Bennett, Stephen Earl. 1988. ““Know-Nothings” Revisited: The Meaning of Political Ignorance Today.” Social Science Quarterly 69 (2): 476–90.
Burns, Peter F, Peter L Francia, and Paul S Herrnson. 2000. “Labor at Work: Union
Campaign Activities and Legislative Payoffs in the US House of Representatives.” Social
Science Quarterly 81 (2): 507–522.
Campbell, Angusu, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. 1960. The
American Voter. Chicago: The University of Chicago Press.
39
Card, David, Thomas Lemieux, and W Craig Riddell. 2004. “Unions and Wage Inequality.”
Journal of Labor Research 25 (4): 519–559.
Clark, Paul, and Marick F Masters. 2001. “Competing Interest Groups and Union Members’
Voting.” Social Science Quarterly 82 (1): 105–116.
Collins, Benjamin. 2012. “Right to Work Laws: Legislative Background and Empirical Research.” Congressional Report for Congress.
Converse, Philip E. 1962. “The Nature of Belief Systems in Mass Publics.” In Ideology and
Discontent, ed. David E. Apter. New York: Free Press pp. 206–261.
Dark, Taylor E. 1999. The Unions and the Democrats: An Enduring Alliance. Cornell
University Press.
Facchini, Giovanni, Anna Maria Mayda, and Prachi Mishra. 2011. “Do interest groups affect
US immigration policy?” Journal of International Economics 85 (1): 114–128.
Feldman, Stanley, and Marco R Steenbergen. 2001. “The Humanitarian Foundation of Public
Support for Social Welfare.” American Journal of Political Science 45 (3): 658–677.
Ferejohn, John. 1990. “Information and the Electoral Process.” In Information and Democratic Processes, ed. John Ferejohn, and James Kukli. University of Illinois Press pp. 3–19.
Fong, Christina. 2001. “Social Preferences, Self-Interest, and the Demand for Redistribution.”
Journal of Public Economics 82 (2): 225–246.
Fordham, Benjamin O, and Katja B Kleinberg. 2012. “How Can Economic Interests Influence
Support for Free Trade?” International Organization 66 (02): 311–328.
Freeman, Richard B, and James L Medoff. 1984. What Do unions Do. New York: Basic
Books.
Gosling, Amanda, and Thomas Lemieux. 2004. “Labor market reforms and changes in wage
inequality in the United Kingdom and the United States.” In Seeking a Premier Economy:
The Economic Effects of British Economic Reforms, 1980-2000. University of Chicago
Press pp. 275–312.
Green, Donald Philip, Bradley Palmquist, and Eric Shickler. 2002. Partisan Hearts and
Minds. New Haven: Yale University Press.
Green, John Clifford. 2007. The Faith Factor: How Religion Influences American Elections.
Greenwood Publishing Group.
Hacker, Jacob S, and Paul Pierson. 2011. Winner-Take-All Politics: How Washington Made
the Rich Richer–and Turned Its Back on the Middle Class. Simon and Schuster.
40
Hacker, Jacob S, Philipp Rehm, and Mark Schlesinger. 2013. “The Insecure American:
Economic Experiences, Financial Worries, and Policy Attitudes.” Perspectives on Politics
11 (1): 23–49.
Hainmueller, Jens, and Michael J Hiscox. 2006. “Learning to Love Globalization: Education
and Individual Attitudes toward International Trade.” International Organization 60 (02):
469–498.
Hainmueller, Jens, Michael J Hiscox, and Yotam Margalit. 2013. “Do Concerns about Labour
Market Competition Shape Attitudes Toward Immigration? New Evidence from U.S.
Workers.”.
Hiscox, Michael J. 2006. “Through a Glass and Darkly: Attitudes toward International Trade
and the Curious Effects of Issue Framing.” International Organization 60 (3): 755–780.
Iversen, Torben, and David Soskice. 2001. “An Asset Theory of Social Policy Preferences.”
American Political Science Review 95 (4): 875–894.
Leighley, Jan E, and Jonathan Nagler. 2007. “Unions, Voter Turnout, and Class Bias in the
US Electorate, 1964–2004.” The Journal of Politics 69 (2): 430–441.
Lupia, Arthur. 1994. “Shortcuts versus Encyclopedias: Information and Voting Behavior in
California Insurance Reform Elections.” American Political Science Review 88 (1): 63–76.
Malhotra, Neil, Yotam Margalit, and Cecilia Hyunjung Mo. 2013. “Economic explanations
for opposition to immigration: Distinguishing between prevalence and conditional impact.”
American Journal of Political Science 57 (2): 391–410.
Mansfield, Edward D, and Diana C Mutz. 2009. “Support for Free Trade: Self-Interest,
Sociotropic Politics, and Out-Group Anxiety.” International Organization 63 (03): 425–
457.
Margalit, Yotam. 2012. “Lost in Globalization: International Economic Integration and the
Sources of Popular Discontent.” International Studies Quarterly 56 (3): 484–500.
Margalit, Yotam. 2013. “Explaining Social Policy Preferences: Evidence from the Great
Recession.” American Political Science Review 107 (1): 80–103.
Masters, Marick F, and John T Delaney. 2005. “Organized Labor’s Political Scorecard.”
Journal of Labor Research 26 (3): 365–392.
Masters, Marick F, and John Thomas Delaney. 1984. “Interunion Variation in Congressional
Campaign Support.” Industrial Relations: A Journal of Economy and Society 23 (3): 410–
416.
Masters, Marick F, and John Thomas Delaney. 1987. “Union Political Activities: A Review
of the Empirical Literature.” Industrial and Labor Relations Review 40: 336–353.
41
Mayda, Anna Maria. 2006. “Who is against Immigration? A Cross-Country Investigation of
Individual Attitudes toward Immigrants.” The Review of Economics and Statistics 88 (3):
510–530.
Mayda, Anna Maria, and Dani Rodrik. 2005. “Why Are Some People (And Countries) More
Protectionist than Others?” European Economic Review 49 (6): 1393–1430.
McClosky, Herbert, and John Zaller. 1984. The American Ethos: Public Attitudes toward
Capitalism and Democracy. Harvard University Press Cambridge, MA.
Neuman, W Russell. 1986. The Paradox of Mass Politics: Knowledge and Opinion in the
American Electorate. Cambridge, Mass.: Harvard University Press.
Neustadtl, Alan. 1990. “Interest-Group PACsmanship: An Analysis of Campaign Contributions, Issue Visibility, and Legislative Impact.” Social Forces 69 (2): 549–564.
Oster, Emily. 2014. “Unobservable Selection and Coefficient Stability: Theory and Validation.” NBER Working Paper No. 19054.
Radcliff, Benjamin, and Patricia Davis. 2000. “Labor Organization and Electoral Participation in Industrial Democracies.” American Journal of Political Science 44 (1): 132–141.
Saltzman, Gregory M. 1987. “Congressional Voting on Labor Issues: The Role of PACs.”
Industrial and Labor Relations Review pp. 163–179.
Scheve, Kenneth F, and Matthew J Slaughter. 2001a. “Labor Market Competition and
Individual Preferences over Immigration Policy.” Review of Economics and Statistics 83
(1): 133–145.
Scheve, Kenneth F, and Matthew J Slaughter. 2001b. “What Determines Individual TradePolicy Preferences?” Journal of International Economics 54 (2): 267–292.
Sears, David O., and Carolyn L. Funk. 1990. “Self-Interest in American’ Political Opinions.”
In Beyond Self-Interest, ed. Jane J. Mansbridge. Chicago and London: The University of
Chicago Press.
Slaughter, Matthew J. 2007. “Globalization and declining unionization in the United States.”
Industrial Relations: A Journal of Economy and Society 46 (2): 329–346.
Smith, Lauren E, and Lee Demetrius Walker. 2013. “Belonging, Believing, and Group Behavior Religiosity and Voting in American Presidential Elections.” Political Research Quarterly
66 (2): 399–413.
Sousa, David J. 1993. “Organized Labor in the Electorate, 1960-1988.” Political Research
Quarterly 46 (4): 741–758.
42
Appendix
1
Data Description
1.1
Survey Items
A number of survey items asked union members about their knowledge regarding their union’s
policy stance and related activities. These items read as follows:
• Overall, where do you think the union stands on the question of whether trade with
other countries should be expanded, reduced or kept at its current level?
• How familiar would you say you are with the union’s view on trade with other countries?
Do you think that trade with other countries is good or bad for you and your family?
• During the past year, approximately how often would you estimate the union has
communicated with you about trade with other countries?
1.2
Calculation of the Union Protectionism Score
The metric of unions’ protectionism score is based on examining the position of the unions
on all trade-related issues that were actively lobbied by at least one of the labor unions in
the study. Given how labor intensive examining each of the fifteen unions’ position on each
of the issues and to make the comparison work managable in scope, we focus only on issues
that were tied to a legislative bill. This focus leaves out issues that were not related to any
legislative activity during the two years period under examination. We also excluded bills
that were purely industry-specific (e.g. the United Steelworkers lobbying on an antidumping
and countervailing duty case on coated paper). Such bills do not allow for a comparison with
the position taken by unions outside the industry since they do not lobby or announce their
position on the issue.
Based on these criteria, we used the fourteen bills in the table below to generate the
unions’ protectionism measure. In the few instances were information was provided only
about the lobbying activity taking place but not about the actual stance taken by the unions,
the position value was coded as missing.
• We examine each union’s position on every specific bill. For instance, we consider if
the union supported or opposed the bill proposing to withdraw from NAFTA or the
43
Table A1: Congressional Bills on Foreign Trade Lobbied by Unions
Lobbied Bills Included for Measuring Union’s Protectionist Stance
Coding
Trade Reform, Accountability, Development and Employment Act of 2009
protectionist
US-Colombia Trade Promotion Agreement
free trade
US-Panama Free Trade Agreement
free trade
US-Korea Free Trade Agreement
free trade
US-Peru Trade Promotion Agreement Implementation Act
free trade
Withdrawal of the US from NAFTA
protectionist
Reauthorizing Trade Adjustment Assistnace
protectionist
Currency Reform for Fair Trade Act; Currency Exchange Rate Oversight Reform Act of 2009 protectionist
Trade Enforcement Act of 2009
protectionist
Buy Amreican Provision in the American Recovery and Reinvestment Act of 2009
protectionist
Reciprocal Market Access Act of 2009
protectionist
U.S. Foreign-Trade Zones: Trade Agreement Parity Proposal
missing info
Trade Agreement Benchmarks and Accountability Act
missing info
Export Promotion Act of 2010
missing info
buy American provision in the American Recovery and Reinvestment Act. Yet, we do
not take into account its general view toward NAFTA or the buy American program.
• We code unions’ support for (opposition to) the trade agreement deals as pro-liberalization
(protectionist), and support for (opposition to) other anti-liberalization bills as protectionist (pro-liberalization).
• Specifically, we code support for the “Trade Reform, Accountability, Development and
Employment Act” of 2009 as a protectionist stance, because the bill: i) Required a
re-evaluation of US free trade agreements every two years; ii) Would have restricted
the applicability of trade agreements in regards to trade in services, foreign investment,
government procurement, IPR protection, trade remedies, among other areas, and iii)
Required the President to submit to Congress a plan to renegotiate any trade agreement
that does not meet the stated requirement already in effect.
• Support for the Trade Enforcement Act of 2009 and the Reciprocal Market Access
Act of 2009 is coded as protectionist. Trade Enforcement Act of 2009 proposes to
apply countervailing duties to non-market economy countries. The Reciprocal Market
Access Act proposes to limit the President’s authority to reduce or eliminate tariffs
pursuant to trade agreements until certain conditions are met, as well as to withdraw
tariff concessions against trade partners who violated the trade agreement.
44
2
Changes in the United States - Korea Free Trade Agreement
The United States and the Republic of Korea initially signed the free trade agreement on
June 30, 2007, but later reached a new agreement that entailed revised provisions for the
automotive sector on December 3, 2010. This sections presents the key issues revised in the
agreement.30
• Tariffs For motor vehicles principally designed for the transport of persons, the United
States shall keep duties at the base rate during years one through four, and eliminate
duties effective January 1 of year five. Korea will reduce duties to four percent ad
valorem on the date KORUS enters into force and eliminate duties effective January 1
of year five.
• Safety Standards “Korea shall provide that an originating motor vehicles of the
United States produced by a manufacturer that sold no more than 25,000 originating motor vehicles in the territory of Korea during the previous calendar year shall
be deemed to comply with Korean Motor Vehicle Safety Standards if the manufacturer certifies that the motor vehicle complies with U.S. Federal Motor Vehicle Safety
Standards.”
• Motor Vehicle Safeguard“Neither Party may apply a safeguard measure for a period
exceeding two years, except that the period may be extended by up to two years if the
competent authorities of the importing Party determine [...] that the measure continues
to be necessary to prevent or remedy serious injury and to facilitate adjustment and
that there is evidence that the industry is adjusting, provided that the total period of
application of a safeguard measure, including the period of initial application and any
extension thereof, shall not exceed four years.”
• Environmental Standards “With regard to Korea’s new automobile fuel economy
and greenhouse gas emissions regulation, [...] Korea will provide that, from 2012 to
2015, a manufacture that sold up to 4500 motor vehicles in the territory of Korea in
calendar year 2009 shall be deemed to comply with the target level set forth in the
regulations if either the average fuel economy or the average CO2 emissions level for the
vehicles the manufacturer sold in the territory of Korea during the relevant calendar
30
The full legal texts are available here:
https://ustr.gov/trade-agreements/free-tradeagreements/korus-fta/legal-texts-reflecting-december-3-2010-agreement.
45
year meets a target level that is 19 percent more lenient than the relevant target level
provided in the regulation that would otherwise be applicable to that manufacturer.”
3
Unobservable Selection and Bounding the Treatment Effect
The method advanced in Oster (2014) shows that we can identify the bounded set of the
treatment effect using the regression values from uncontrolled and controlled regressions and
assumptions about: (i) δ̄, the proportional selection between observables and unobservables
related to the treatment and (ii ) Rmax , the R-squared of the full regression with the treatment, observable controls and unobservable controls.
The δ̄ captures the relative importance of the index of observed and unobserved variables
in explaining the treatment. The bound δ̄ = 1 means that the unobservables are as important
as the observables. This is considered an appropriate upper bound because “researchers
typically focus their data collection efforts (or their choice of regression controls) on the
controls they believe ex ante are the most important” (Oster, 2014: 11). It is thus quite
unlikely that unobservables are more important than the whole set of observable controls
that are relevant to the treatment. Regarding Rmax , it is necessarily bounded between R̄, the
R-squared of the controlled regression, and 1. The simulated results suggest that the upper
bound of min{2.2R̄, 1} is an appropriate assumption to make. We therefore calculated the
lower bound of the “union effect” assuming δ̄ = 1 and 2.2R̄.
46